Forums  > Pricing & Modelling  > Multi horizon forecasting  
     
Page 1 of 1
Display using:  

cricketLover192


Total Posts: 2
Joined: Jun 2020
 
Posted: 2020-06-05 00:38
Hello, I am looking to better understand methodologies for forecasting models on different horizon. For example, I find predictive power at t+2 day forward time, and also model at t+1 day forward time, and t+1 hour forward time. These are varying strength and goodness -- % variance, etc.

The predictive power is not always good for whole period and may even be negative in shorter horizons, so it is not clear to me I should use pro-ration or similar technique to try to apply longer into shorter horizon and iterate at each horizon period.

Is any good literature or industry standard approaches to this type of problem? I am trying to use for building portfolios and rebalancing them at whatever is optimal time points. However, it is seeming to be difficult to capture model quality. Is suggestion to use independent tests on multiple horizons and some kind of joint objective on this?

Thanks, very much:-

rickyvic


Total Posts: 217
Joined: Jul 2013
 
Posted: 2020-06-07 20:43
A Comparison of Direct and Iterated Multistep AR Methods for Forecasting Macroeconomic Time Series, Marcellino, Stock and Watson 2006

Then use that knowledge for your application, there are a lot of advancements but the idea is in there.
If you have question contact Marcellino he is very nice.

"amicus Plato sed magis amica Veritas"
Previous Thread :: Next Thread 
Page 1 of 1